In untreated patients with lung, chest wall, or control abnormalities, blood gases typically deteriorate somewhat in NREM sleep, and then deteriorate much further in REM sleep. This deterioration is likely due to multiple causes, including:                1. Increased upper airway resistance due to pharyngeal collapse.        2. Loss of cough and sigh, leading to sputum retention and atelectasis.        3. Postural effects on V/Q.        4. Reduced tonic or chemoreflex drive to the diaphragm, particularly in REM.        5. Reduced tonic or chemoreflex drive to intercostals, abdominal expiratory muscles, and other accessory muscles.        6. Possible REM-specific changes in pulmonary blood flow distribution.Pharyngeal collapse is most profound in REM sleep. There is evidence that the reduction in ventilation in NREM sleep is entirely due to pharyngeal collapse, and not to a reduction in chemoreflex drive to the diaphragm. Increased pharyngeal resistance is treated with CPAP, or more generally with positive pressure sufficient to splint the airway at zero flow, plus additional inspiratory pressure sufficient to compensate for resistive and Bernoulli pressure drop.        
Reducing the work of breathing and resting the respiratory muscles by providing ventilatory support, particularly if delivered during sleep, can have a number of direct and indirect potential benefits. These benefits include:                Prevention of muscle fatigue with inefficient contraction.        Reduced oxygen cost of breathing.        Reduction of dyspnea.        Improved sleep, with fewer respiratory arousals.Improved sleep should in turn reduce metabolic rate, CO2 production and oxygen consumption, directly and indirectly by reduced rolling around, fidgeting, etc., leading to either better blood gases or reduced need for ventilatory support. It is also worthwhile in its own right because of improved quality of life.        
However, there are some untoward effects of ventilatory support on the patient as follows:
1. Barotrauma
For ventilators delivering less than 35 cmH2O peak pressure, barotrauma is largely confined to patients with adult respiratory distress syndrome (due to high shear stresses) and to patients with a history of pneumothorax or emphysematous bullae.
2. Reduced Cardiac Output
Even in normal subjects, 10 cmH2O nasal CPAP can produce a 10% reduction in cardiac output, and high levels of positive pressure, particularly in patients who are fluid depleted, can produce a profound reduction in cardiac output. Conversely, in patients with cardiac failure and fluid overload (pulmonary capillary wedge pressure in excess of 15 cmH2O), nasal CPAP actually increases cardiac output, probably by reducing transmural pressure.
3. Mouth Leak
Mouth leak is present to some extent in most patients being treated with ventilatory support. A mouth leak of 0.4 L/sec causes severe sleep disruption, loss of ventilatory support, loss of supplemental oxygen, and loss of end expiratory splinting pressure. Such a leak is present in perhaps 50% of subjects. Mouth leak also causes increased nasal resistance. This is a reflex response to drying and cooling of the nasal mucosa by a unidirectional flow of air in the nose and out the mouth.
A chin strap is only very partially effective in controlling mouth leak. Heated humidification can partially treat the, drying of the nasal mucosa but not the other aspects of the problem. Where tolerated, a full face mask is the preferred treatment.
4. Glonic Closure
Rodenstein and colleagues have shown that over ventilation leads to a progressively tight closure of the vocal cords, both awake and asleep, and that this fact must be taken into account when providing noninvasive ventilation.
The details are not well understood; it is not known whether the glottic closure is purely passive or involves active adduction, whether it is abolished by anaesthesia, whether it is present in REM, whether it is due to airway or arterial hypocapnia, or whether it is produced by sleepstate specific changes in set-point. Unlike passive pharyngeal collapse, it is not known whether vocal cord closure responds to CPAP, but if it is an active closure it would be expected to be extremely refractory to CPAP.
5. Increased Deadspace
Positive pressure will alter the distribution of pulmonary blood flow, tending to reduce blood flow to poorly ventilated units (beneficial reduction in physiological shunt) and also to well-ventilated units (pathological increase in deadspace). In patients in whom there is much blood flow to poorly perfused lung units, for example patients with obesity hypoventilation syndrome, this reduction in physiological shunt but increase in deadspace can be of net benefit, whereas in patients with much ventilation to poorly perfused regions, such as “pink puffers”, the net effect can be detrimental.
6. Discomfort
A goal of a ventilator is to relieve dyspnea. However, it can cause considerable discomfort, by various mechanisms:                Distension of upper airway structures.        Swallowing of air (particularly once pressures exceed 20 cmH2O)        Mask discomfort.        Leak, particularly mouth leak.        Patient-machine asynchrony.        
We might expect that as the degree of support is increased from zero towards that which will perform 100% of eupneic respiratory work, the sense of dyspnea due to having to do an abnormally high amount of respiratory work, and the sense of distress due to excess chemoreflex stimulation should both decrease towards zero. However, discomfort from all the causes bulleted above will increase. There is no literature on the rate of trade-off between the two sources of distress, but it is apparent that the patient should feel most comfortable at a degree of support which is less than 100% support. Very preliminary unpublished work by the current author, in which normal subjects breathe through a high external resistance (8 cmH2O/L/sec) with 200 ml added deadspace, and are then treated with bilevel support, the patient feels most comfortable at about 50% support. The optimum point may of course be quite different in a patient with actual lung or chest wall disease, or with forms of support other than bilevel.
7. Patient-Machine Asynchrony
Patient-machine asynchrony can be due to a number of factors, including:                Leaks.        Long respiratory time constant (e.g. in patients with severe chronic airflow limitation (“CAL”).        Intrinsic PEEP.        
Leaks, and particularly variable leaks, cause asynchrony because the airflow measured by the device does not equal the patient respiratory airflow. With a device of the invention, leaks start to become a problem at about 0.2 L/sec, and are a severe problem by 0.4 L/sec. At 0.6 L/sec, the device will probably not really be benefiting the patient. Keeping the leak much below 0.2 L/sec is technically very demanding and not generally practicable. Therefore, while one wants to keep the leak as low as possible with reasonable investment of effort, 0.2 L/sec is a reasonable balance between effort and results.
Patient-machine asynchrony is particularly a problem in patients with long respiratory time constants being treated with high degrees of support. This is because even true respiratory airflow no longer equals patient effort. For example, at the end of the patient's inspiratory effort, the lungs have not yet equilibrated to the high inspiratory pressure and continue to fill. This prevents correct triggering into expiration. The patient must actively expire in order to terminate the inspiration. The higher the degree of support results in greater difficulty with the phenomenon. Therefore, one wants to avoid excessive support.
Intrinsic PEEP causes a kind of asynchrony because the patient must generate a considerable inspiratory effort before any flow is generated. Intrinsic PEEP due to dynamic airway compression may be evident from an expiratory flow-time curve, in which there is a brief period of very high expiratory flow, followed by a very prolonged expiratory flow plateau at a much lower flow. Treatment is to increase expiratory pressure (particularly late expiratory pressure) until the curve shape normalizes.
Thus, with these seven effects in mind, the goals of automatic ventilatory positive airway pressure may generally be summarized to include the following:                1. To guarantee an adequate alveolar ventilation during sleep.        2. To maximize wake comfort.        3. To maximize depth of sleep.        4. To minimize cost of initiation of therapy.Directed towards the above goals, a ventilator device in accordance with the invention may provide:        1. Servo-control of minute ventilation to equal or exceed a chosen target.        2. Unloading of much of the spontaneous resistive work if the subject exceeds the chosen target.        3. A smooth and physiological pressure waveform whose minimum amplitude will unload much but not all of spontaneous elastic work if the subject just exceeds the chosen target.        4. A mechanism for automatically establishing the target during an awake learning session in subjects who have adequate PCO2 in the daytime and who deteriorate only during sleep.        
However, even sophisticated ventilatory devices with a high degree of automatic processing developed to meet one or more of these goals such as the devices disclosed in International Publication No. WO 98/12965 and International Publication No. WO 99/61088 still often require the setting of controls to accommodate a particular patient's needs before beginning use. Absent a uniform methodology for adjusting the settings of such a device, the delivery of the appropriate degree of pressure support to the patient may not be optimal.